Coordinated Optimization Control System for Deep Peak Shaving and Primary Frequency Regulation of Thermal Power Units Based on Multimodal Adaptive Learning

Abstract

This invention discloses a multimodal adaptive learning-based collaborative optimization control system for deep peak shaving and primary frequency regulation in thermal power units, belonging to the field of intelligent control technology for thermal power units. The system includes a multi-source operating data sensing module, an operating mode recognition module, a hierarchical collaborative optimization decision-making module, a collaborative execution module, a safety protection module, and a digital twin support platform. By constructing a closed-loop control architecture of "sensing—modal recognition—hierarchical decision-making—collaborative execution," the system dynamically adjusts the priority of control objectives according to the unit's operating modes, achieving collaborative optimization control of deep peak shaving operation and primary frequency regulation response. During the deep peak shaving phase, the system constructs controllable steam energy storage reserves; during the frequency regulation response phase, it employs a coordinated strategy of rapid turbine regulation and boiler compensation to improve the unit's frequency response capability under low-load conditions and reduce main steam pressure fluctuations. The digital twin support platform participates in online control decision-making, improving system stability and adaptability.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for thermal power generating units, specifically relating to a deep peak shaving and primary frequency regulation coordinated optimization control system for thermal power generating units based on multimodal adaptive learning. It is particularly suitable for intelligent control systems that achieve coordinated optimization of deep peak shaving control and primary frequency regulation control under wide load operation conditions of thermal power generating units when a high proportion of new energy sources are connected to the grid. Background Technology

[0002] In recent years, driven by energy structure transformation and the "dual carbon" target, the proportion of renewable energy sources such as wind power and photovoltaics in the power system has continued to increase. Due to the randomness, volatility and intermittency of new energy power generation, grid frequency stability faces greater challenges. Traditional thermal power units are gradually transforming from baseload power sources to flexible regulation power sources, undertaking ancillary service functions such as deep peak shaving and primary frequency regulation in the power system.

[0003] Currently, thermal power units typically employ a coordinated control system (CCS) to jointly control the boiler and turbine. Common control modes include boiler-following-turbine, turbine-following-boiler, and direct energy balance control strategies. These control systems maintain good stability and economy when operating at rated load or in medium-to-high load ranges. However, when the unit load drops to 30% of rated load or even lower in the deep peak-shaving range, boiler combustion stability decreases, main steam pressure and temperature fluctuations increase, turbine energy storage capacity weakens, and the system's dynamic characteristics change significantly. Traditional fixed-parameter control strategies struggle to balance stability and rapid response performance.

[0004] On the other hand, primary frequency regulation is a dynamic response process on the order of seconds, requiring generating units to rapidly adjust their active power output when the grid frequency deviates. Deep peak shaving, however, is a load regulation process on the order of minutes or even hours, with control objectives typically focusing on economy and safety. Due to the differences in time scale, energy storage methods, and control objectives between the two types of control tasks, the primary frequency regulation capability of generating units often decreases significantly under deep peak shaving conditions. This manifests as prolonged response lag time, insufficient regulation capacity, and severe fluctuations in main steam pressure, making it difficult to meet the grid's requirements for rapid frequency support.

[0005] Furthermore, the boiler-turbine system is a highly nonlinear, multivariable, strongly coupled, large-inertial object. Parameters such as main steam pressure, main steam temperature, reheat temperature, furnace negative pressure, oxygen content, and NOx emissions are interrelated and affected by factors such as fuel characteristics, equipment aging, heat exchanger fouling, and changes in ambient temperature. When operating over a wide load range, system parameters exhibit significant condition-dependent characteristics, making it difficult for traditional control methods based on fixed or linearized models to maintain good control quality across all operating ranges.

[0006] To address the aforementioned issues, existing technologies have proposed some improvement measures. For example, these include improving primary frequency regulation performance by optimizing the droop coefficient, adding feedforward compensation, and introducing energy storage compensation control logic; or enhancing the unit's dynamic response capability through advanced control methods such as model predictive control and adaptive control. However, these methods are mostly local optimizations targeting a single control objective, lacking a unified and coordinated design for deep peak shaving and primary frequency regulation at the system level. Furthermore, most advanced control strategies are highly dependent on model accuracy, leading to model mismatch, complex parameter tuning, and high operational and maintenance difficulties in practical engineering applications.

[0007] Therefore, with the high proportion of renewable energy integration, the regulation tasks undertaken by thermal power units are becoming increasingly complex, exhibiting characteristics of wide load, multiple operating conditions, and strong dynamic disturbances. There is an urgent need for a collaborative optimization control system capable of adapting to wide load operating conditions, dynamically adjusting control objectives at different operating stages, balancing deep peak shaving and primary frequency regulation performance, and simultaneously possessing adaptive regulation capabilities and operational safety assurance mechanisms. This system would address technical problems in existing technologies such as the mutual constraints between control objectives of deep peak shaving and primary frequency regulation, the difficulty in coordinating multi-variable coupling relationships, and the insufficient adaptability of control strategies to changes in operating conditions.

[0008] Furthermore, existing control systems are typically designed based on fixed control modes or a single optimization objective, without dynamically reconfiguring the weights of the control objective according to the actual operating modes of the unit, nor actively building controllable steam energy storage reserves during the deep peak shaving phase to support subsequent primary frequency regulation response. This results in a decrease in the unit's frequency regulation capability under low load operating conditions, making it difficult to achieve unified optimization of peak shaving and frequency regulation. Summary of the Invention

[0009] In view of this, the purpose of this invention is to provide a multimodal adaptive learning-based deep peak shaving and primary frequency regulation coordinated optimization control system for thermal power units, so as to solve the problems of mutual constraints between deep peak shaving and primary frequency regulation control objectives, difficulty in coordinating strong coupling of multiple variables, and insufficient adaptability of control strategies to wide load operating conditions in the prior art, and improve the regulation performance and operational stability of the unit under the background of high proportion of new energy access.

[0010] To achieve the above objectives, the present invention provides the following technical solution: In one embodiment of the present invention, a multimodal adaptive learning-based deep peak shaving and primary frequency regulation coordinated optimization control system for thermal power units is provided, including a multi-source operation data sensing module, an operation mode recognition module, a hierarchical coordinated optimization decision module, a coordinated execution module, and a digital twin support platform; wherein, the control system adopts a closed-loop architecture of "sensing-modal recognition-hierarchical decision-coordinated execution", and achieves coordinated optimization of deep peak shaving operation and primary frequency regulation response by switching control targets driven by operation modes.

[0011] In one embodiment of the present invention, the multi-source operation data sensing module is used to collect the operating parameters of the boiler system, turbine system and generator system of the thermal power unit, as well as the power grid frequency signal and dispatch command signal, thereby constructing a basic dataset of the unit's operating status.

[0012] Furthermore, the operating mode identification module is used to analyze the unit's operating status based on the operating parameters and identify the unit's current operating mode. The operating mode includes at least the deep peak shaving steady-state mode, the deep peak shaving transition mode, the primary frequency regulation response mode, and the conventional load operating mode.

[0013] Preferably, the operating mode identification module classifies the operating state based on the unit load level, main steam pressure, main steam temperature, grid frequency deviation, frequency change rate, and load change rate, and outputs the current operating mode and mode confidence level through cluster analysis combined with time series feature identification method.

[0014] Furthermore, the hierarchical collaborative optimization decision module includes a target weight dynamic adjustment unit and an adaptive control unit; the target weight dynamic adjustment unit is used to adjust the priority of the control target according to the current operating mode, and the adaptive control unit is used to generate coordinated control commands for the boiler and the steam turbine.

[0015] Preferably, the target weight dynamic adjustment unit adopts different control target priority configurations under different operating modes; under the deep peak shaving steady-state mode, economical operation and frequency regulation capacity maintenance are the priority targets; under the primary frequency regulation response mode, the frequency fast response and response accuracy are the priority targets; under the deep peak shaving transition mode, the stability of operating parameters and equipment safety are the priority targets.

[0016] Furthermore, the adaptive control unit adopts a control method that combines a mechanistic model and a data-driven model. The mechanistic model is used to handle the multivariate coupling relationship and operating constraints of the boiler-turbine system, while the data-driven model is a prediction model or strategy optimization model trained based on historical operating data. It is used to predict the dynamic response of the unit or to compensate for the control output, so as to improve the control accuracy and stability over a wide load range.

[0017] Optionally, under deep peak shaving operation, the hierarchical collaborative optimization decision module optimizes the turbine valve opening and appropriately increases the main steam pressure setpoint to form a controllable steam energy storage reserve, thereby enhancing the unit's rapid power response capability during primary frequency regulation.

[0018] Furthermore, during a frequency regulation response process, the hierarchical collaborative optimization decision module adopts a control strategy that prioritizes rapid adjustment of the turbine control valve, supplements it with boiler energy storage compensation, and coordinates the follow-up of fuel and air volume, in order to reduce main steam pressure fluctuations and reduce reverse coupling regulation between the boiler and the turbine.

[0019] In one embodiment of the present invention, a safety protection module is also included, which is used to monitor in real time the margin between the main steam pressure, main steam temperature, furnace negative pressure and other key operating parameters and the preset operating boundary.

[0020] Furthermore, when the key operating parameters approach the preset operating boundary, the current optimization control target is paused and a safety-priority control strategy is initiated to ensure the safe operation of the unit.

[0021] Optionally, the digital twin support platform is used to construct a unit mechanism model and a data fusion model, and to provide model parameters, operating boundary constraint information or dynamic prediction results to the hierarchical collaborative optimization decision-making module to assist in online control decision-making.

[0022] Based on the above technical solution, the multimodal adaptive learning-based thermal power unit deep peak shaving and primary frequency regulation coordinated optimization control system of the present invention forms a dynamic adaptive control mechanism throughout the entire operation process of the unit by constructing a closed-loop control architecture of "perception-modal recognition-hierarchical decision-coordinated execution". By reconstructing the weight of the control target driven by the operating mode and switching the coordinated control strategy, the system achieves coordination and unification between deep peak shaving and primary frequency regulation, thereby improving the frequency response capability of the unit under low load operating conditions, shortening the frequency regulation response time, and reducing the main steam pressure fluctuation amplitude.

[0023] Specifically, this invention uses a multi-source operation data sensing module to collect key operating parameters of the boiler system, turbine system, and generator system, as well as grid frequency signals, in real time to establish a global perception foundation for the unit's operating status. The operating mode recognition module classifies and judges the current operating status of the unit, enabling the control system to identify different operating scenarios such as deep peak shaving steady state, transient state, and primary frequency regulation response state, thereby avoiding the problem of control quality degradation in traditional fixed control mode over a wide load range.

[0024] Based on this, the hierarchical collaborative optimization decision module dynamically adjusts the priority of control objectives according to the operating modes, enabling the control system to automatically switch the control focus under different operating conditions: prioritizing economical operation and frequency regulation capability reserves under deep peak shaving steady state; prioritizing rapid frequency response and response accuracy during primary frequency regulation response; and prioritizing parameter stability and equipment safety during load transition phases. By dynamically reconstructing the control objectives driven by the operating modes, the conflict between deep peak shaving and primary frequency regulation control objectives in existing technologies is resolved.

[0025] Meanwhile, this invention improves control accuracy and adaptability to nonlinear conditions by combining mechanistic models with data-driven models in an adaptive control mode, while ensuring the stability of multivariable coupling in the boiler-turbine system, thus enabling the unit to maintain stable operation over a wide load range.

[0026] Furthermore, under deep peak-shaving operation, by optimizing the turbine valve opening and appropriately increasing the main steam pressure setpoint, a controllable steam energy storage reserve is formed, providing rapid power release capability for primary frequency regulation. During the primary frequency regulation response, a control strategy is adopted, with rapid turbine valve adjustment as the primary method, boiler energy storage compensation as the secondary method, and fuel and air volume coordinated follow-up, to reduce main steam pressure fluctuations and reduce the reverse coupling regulation phenomenon between the boiler and turbine, thereby improving the frequency regulation response speed and stability.

[0027] Furthermore, this invention establishes a digital twin support platform, which, by constructing a fusion model of unit mechanism and data, provides model parameters, operational boundary constraint information, or dynamic prediction results to the hierarchical collaborative optimization decision-making module, enabling the control system to have prediction and correction capabilities during online operation, thereby improving the reliability and security of the control strategy.

[0028] Meanwhile, the safety protection module monitors the boundaries of key operating parameters. When the parameters approach the preset operating boundaries, the safety priority control strategy is automatically activated to ensure the safe operation of the unit while ensuring the optimized control effect.

[0029] Therefore, while achieving deep peak shaving operation, this invention effectively improves the primary frequency regulation response capability and operational stability of the unit, alleviates the inherent contradiction between deep peak shaving and primary frequency regulation, enhances the unit's economy, safety and grid support capability over a wide load range, and realizes the transformation from traditional fixed-mode control to multi-modal adaptive collaborative optimization control. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall structure of the multimodal adaptive learning-based deep peak shaving and primary frequency regulation collaborative optimization control system for thermal power units according to the present invention. Figure 2 This is a schematic diagram of the hierarchical collaborative optimization decision module of the present invention; Figure 3 This is a logical schematic diagram of the deep peak modulation and primary frequency modulation coordinated control strategy of the present invention; Figure 4 This is a schematic diagram illustrating the relationship between the digital twin support platform and the online control system of this invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the described embodiments are only used to explain the technical solutions of this invention and are not intended to limit the scope of protection of this invention. Without departing from the spirit and essence of this invention, those skilled in the art can make various equivalent substitutions or modifications to the technical solutions of this invention, and all such equivalent substitutions or modifications should fall within the scope of protection of this invention.

[0032] Example 1: System Overall Architecture To make the technical solution of the present invention clearer, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the described embodiments are only for illustrating the present invention and are not intended to limit the scope of protection of the present invention.

[0033] like Figure 1 As shown, the present invention provides a multimodal adaptive learning-based deep peak shaving and primary frequency regulation collaborative optimization control system for thermal power units. The overall system adopts a closed-loop control architecture of "sensing-modal recognition-hierarchical decision-making-collaborative execution", mainly including: a multi-source operation data sensing module 10, an operation modality recognition module 20, a hierarchical collaborative optimization decision-making module 30, a collaborative execution module 40, a safety protection module 50, and a digital twin support platform 60. The modules interact bidirectionally through data interfaces and control interfaces to achieve dynamic adaptive control throughout the entire operation of the unit.

[0034] I. Applicable Users of the System The control system described in this embodiment is applicable to thermal power generating units, especially coal-fired power units undertaking deep peak shaving and primary frequency regulation tasks in an environment where a high proportion of new energy sources are integrated into the grid. The unit can be a subcritical unit, a supercritical unit, or an ultra-supercritical unit, with a capacity range of 300MW and above, and can also be expanded to other capacity levels as needed.

[0035] II. Overall Functional Architecture 1. Multi-source operational data sensing module 10 The multi-source operation data sensing module 10 is used to collect the operating parameters of the boiler system, turbine system and generator system of the thermal power unit in real time, and to collect the power grid frequency signal and dispatch command signal.

[0036] The operating parameters include, but are not limited to: unit load, main steam pressure, main steam temperature, reheat steam temperature, feedwater flow rate, fuel quantity, air volume, furnace negative pressure, and turbine control valve opening.

[0037] The power grid signal includes frequency deviation and frequency change rate.

[0038] The scheduling command signals include AGC commands and load adjustment commands.

[0039] Through the above data collection, a comprehensive data foundation for the unit's operating status is constructed.

[0040] 2. Run the modal recognition module 20 The operating mode recognition module 20 is connected to the multi-source operating data sensing module 10 to perform real-time analysis of the unit's operating status and output the current operating mode label and mode confidence level.

[0041] The operating modes include at least: Deep peak shaving steady-state mode Deep peaking transition mode First frequency modulation response mode Normal load operation mode The results of operational mode recognition are used as criteria for subsequent control target switching.

[0042] 3. Hierarchical Collaborative Optimization Decision Module 30 The hierarchical collaborative optimization decision module 30 is connected to the operation mode identification module 20. It is used to dynamically adjust the priority of control targets according to the current operation mode and the power grid frequency deviation signal, and generate coordinated control commands for the boiler and the steam turbine.

[0043] This module includes a target weight dynamic adjustment unit 31 and an adaptive control unit 32, the specific structure of which will be described in detail in subsequent embodiments.

[0044] 4. Collaborative Execution Module 40 The collaborative execution module 40 is connected to the hierarchical collaborative optimization decision module 30, and is used to apply the coordinated control commands to: Fuel regulation circuit Air volume regulation circuit Water supply regulating circuit Steam turbine regulating valve circuit Through the above-mentioned multi-loop coordinated regulation, the joint control of the boiler-turbine system is realized.

[0045] 5. Safety protection module 50 The safety protection module 50 and the collaborative execution module 40 are connected in parallel to monitor the margin between key operating parameters such as main steam pressure, main steam temperature, and furnace negative pressure and the preset operating boundary in real time.

[0046] When a critical parameter is detected to be close to a preset boundary, the safety protection module 50 suspends the current optimized control target and initiates a safety priority control strategy to ensure the safe operation of the unit.

[0047] 6. Digital Twin Support Platform 60 The digital twin support platform 60 interacts with the hierarchical collaborative optimization decision-making module 30.

[0048] The digital twin support platform 60 is used to construct unit mechanism models and data fusion models, and provides them to the hierarchical collaborative optimization decision module 30. Model parameters Run boundary constraint information Dynamic prediction results This assists in online control decision-making and improves the control system's adaptability to complex operating conditions.

[0049] III. Data and Control Flow In this embodiment, the basic operation flow between the modules is as follows: 1. The multi-source operation data sensing module 10 collects unit operation data and grid signals in real time; 2. The operating mode recognition module 20 identifies the current operating mode based on the data; 3. The hierarchical collaborative optimization decision-making module 30 adjusts the priority of control objectives and generates coordinated control commands based on the operating mode; 4. The collaborative execution module 40 executes control instructions; 5. The security protection module 50 performs boundary monitoring on the execution results; 6. The digital twin support platform 60 provides prediction and model support for the hierarchical collaborative optimization decision-making module 30, enabling the optimization of control strategies.

[0050] Through the above closed-loop process, the unit can automatically adjust the control strategy under different operating conditions, and achieve synergistic optimization between deep peak shaving operation and primary frequency regulation response.

[0051] Example 2: Implementation of Multimodal Recognition like Figure 1 As shown, the operating mode recognition module 20 is connected to the multi-source operating data sensing module 10. It is used to perform real-time analysis and classification of the current operating status of the unit, and output the operating mode label and mode confidence level to provide decision-making basis for the hierarchical collaborative optimization decision module 30.

[0052] I. Input Data Composition In this embodiment, the operational modality recognition module 20 receives real-time data from the multi-source operational data sensing module 10, and the input data includes, but is not limited to: 1. Unit load values ​​and load change rate; 2. Main steam pressure and pressure change rate; 3. Main steam temperature; 4. Power grid frequency deviation and rate of frequency change; 5. Turbine control valve opening; 6. Boiler fuel quantity and air volume regulation status.

[0053] The above parameters can reflect the current load range, dynamic change trend, and grid frequency disturbance status of the unit, and are the main basis for judging the operating mode.

[0054] II. Principles of Operational Mode Division In this embodiment, the operating mode identification is based on a comprehensive judgment of load level and frequency disturbance state, and the specific classification principle is as follows: 1. Normal load operation mode: When the unit load is in the medium-high load range of the rated load and the grid frequency deviation is within the set stable range, it is determined to be the normal load operation mode.

[0055] 2. Deep peak shaving steady-state mode: When the unit load drops to the preset deep peak shaving range and the load change rate is lower than the set threshold, and the grid frequency does not trigger the frequency regulation response condition, it is determined to be deep peak shaving steady-state mode.

[0056] 3. Deep peak shaving transition mode: When the unit load is within the deep peak shaving range and the load change rate is higher than the set threshold, it is determined to be in the deep peak shaving transition mode.

[0057] 4. Primary frequency regulation response mode: When the detected grid frequency deviation exceeds the preset frequency regulation trigger threshold, or the frequency change rate exceeds the set value, it is determined to be the primary frequency regulation response mode.

[0058] By comprehensively judging the above multiple conditions, the system can distinguish different operating scenarios, thereby providing an accurate basis for subsequent control target switching.

[0059] III. Modal Recognition Implementation Methods In this embodiment, the modal recognition module 20 can be implemented using a method based on feature clustering analysis combined with temporal feature recognition.

[0060] Specifically: 1. First, standardize the input parameters to eliminate the influence of dimensions; 2. Construct a feature vector based on load level, pressure change trend, and frequency deviation; 3. Classify and train historical operational data using cluster analysis to form modal feature centers; 4. During online operation, the real-time feature vector is matched with the preset modal features, and the current modal label is output; 5. Modal confidence is obtained by calculating the distance or matching degree between the current feature and each modality center.

[0061] Preferably, to avoid the control target jitter caused by frequent mode switching, a hysteresis interval or time holding mechanism is set during the mode switching process, that is, the mode switching can be completed only after the recognition result continuously meets the new mode conditions for a preset time.

[0062] IV. Modal Recognition Output Results The output of the modality recognition module 20 includes: 1. Current running modal label; 2. Modal confidence score; 3. Mode switching signal.

[0063] The output result is sent to the hierarchical collaborative optimization decision module 30 to trigger the dynamic adjustment of the control target weight.

[0064] Through the aforementioned modal recognition mechanism, the control system can adaptively switch modes according to the actual operating status of the unit, thereby laying the foundation for the coordinated optimization control of deep peak shaving and primary frequency regulation.

[0065] Example 3: Hierarchical Decision Structure like Figure 2 As shown, the hierarchical collaborative optimization decision module 30 is the core module for realizing the collaborative optimization control of deep peak shaving and primary frequency regulation in this invention. It is connected to the operating mode identification module 20 and the collaborative execution module 40, and is used to dynamically reconstruct the control target according to the current operating mode and generate coordinated control commands for the boiler and turbine.

[0066] In this embodiment, the hierarchical collaborative optimization decision module 30 includes a target weight dynamic adjustment unit 31 and an adaptive control unit 32, which together form a hierarchical decision structure.

[0067] I. Target Weight Dynamic Adjustment Unit 31 The target weight dynamic adjustment unit 31 is used to receive the operating mode label and mode confidence level output by the operating mode recognition module 20, and adjust the priority combination of control targets according to different operating modes.

[0068] (1) Control target type In this embodiment, the control objective includes at least: 1. Economic objectives, aimed at reducing coal consumption for power generation and improving operational efficiency; 2. Frequency regulation performance targets, used to improve the unit's response speed and regulation accuracy to grid frequency deviations; 3. Safety and stability targets are used to keep key parameters such as main steam pressure and main steam temperature within the allowable range.

[0069] (2) Target configuration under different modes Under the deep peak shaving steady-state mode, the target weight dynamic adjustment unit 31 sets the economic target and the frequency regulation capacity maintenance target as priority targets, while ensuring that the safety and stability targets meet the operating boundary requirements.

[0070] In a single frequency modulation response mode, the target weight dynamic adjustment unit 31 sets the frequency modulation performance target to the highest priority and appropriately reduces the weight of the economic target to ensure a fast frequency response.

[0071] Under the deep peak shaving transition mode, the target weight dynamic adjustment unit 31 prioritizes the safety and stability target, controls the rate of parameter change, and avoids oscillation of the unit during rapid load changes.

[0072] By using the dynamic weight allocation method described above, the control system can automatically adjust the control center of gravity under different operating conditions to achieve a balance between multiple objectives.

[0073] II. Adaptive Control Unit 32 The adaptive control unit 32 is used to dynamically adjust the priority combination output by the target weight unit 31 to generate coordinated control commands for the boiler and the steam turbine.

[0074] (1) Control structure In this embodiment, the adaptive control unit 32 adopts a control method that combines a mechanistic model and a data-driven model.

[0075] in: Mechanistic models are used to describe the thermodynamic relationships and multivariable coupling characteristics of boiler-turbine systems, and to handle the inherent constraints between main steam pressure, main steam temperature, fuel quantity and control valve opening. Data-driven models are used to build dynamic prediction models or control compensation models based on historical operating data, so as to improve the controller's ability to adapt to nonlinear characteristics over a wide load range.

[0076] (2) Coordinated control output The control commands output by the adaptive control unit 32 include: Fuel quantity adjustment command; Air volume adjustment command; Water supply flow rate adjustment command; Steam turbine control valve opening adjustment command.

[0077] The aforementioned control commands are applied to the corresponding regulating loops via the collaborative execution module 40, thereby achieving coordinated control between the boiler and the steam turbine.

[0078] (3) Control constraint processing In this embodiment, the adaptive control unit 32 considers the following constraints when generating control commands: 1. Main steam pressure and upper and lower limits of main steam temperature; 2. Combustion stability constraints; 3. Valve opening range constraints; 4. Load change rate constraint.

[0079] By comprehensively processing the above constraints, we ensure that the control commands meet the target weight requirements without exceeding the safe operation boundaries of the unit.

[0080] In this embodiment, in order to achieve dynamic reconstruction of control objectives under different operating modes, the hierarchical collaborative optimization decision module 30 constructs a multi-objective weighted optimization function to uniformly describe the comprehensive optimization relationship between economic objectives, frequency regulation performance objectives, and safety and stability objectives.

[0081] Specifically, the comprehensive optimization objective function is set as follows: J=w1(m)·J1+w2(m)·J2+w3(m)·J 3； in: J is the comprehensive optimization objective function; J1 is the economic objective function, used to characterize the coal consumption for power supply or energy utilization efficiency; J2 is the frequency modulation performance objective function, used to characterize the active power response speed or frequency deviation suppression capability; J3 is the safety and stability objective function, used to characterize the degree of deviation of key parameters such as main steam pressure and main steam temperature; m represents the current operating mode; W1(m), w2(m), and w3(m) are dynamic weight functions related to the operating mode.

[0082] Under different operating modes, the dynamic weight function wᵢ(m) automatically adjusts its value according to the mode type, thereby changing the priority of each control objective in the comprehensive optimization.

[0083] For example: Under deep peak-shaving steady-state mode, the value of w1(m) is higher than that of w2(m) to prioritize economical operation while maintaining a certain reserve of frequency regulation capability. In the first frequency modulation response mode, the value of w2(m) is higher than that of w1(m) to prioritize the fast frequency response performance; In the deep peak-shaving transition mode, the value of w3(m) is increased to enhance operational stability control.

[0084] By introducing the operating mode into the weight function, the weight of the control target becomes a function of the operating mode, thereby achieving dynamic reconstruction of the control target, rather than a fixed weight allocation method.

[0085] Based on this, the adaptive control unit 32 solves the comprehensive optimization objective function J under the premise of satisfying the operating constraints, and generates coordinated control commands for the boiler and the steam turbine.

[0086] III. Technical Effects of Layered Structure The hierarchical decision-making structure formed by the target weight dynamic adjustment unit 31 and the adaptive control unit 32 enables the control system to achieve the following technical effects: 1. Capable of automatically reconfiguring control objectives based on operational modes; 2. It can maintain control stability over a wide load range; 3. It can alleviate the control conflict between deep peak shaving and primary frequency regulation; 4. It can achieve coordinated optimization among control objectives at different time scales.

[0087] Through the aforementioned hierarchical decision-making mechanism, the control system of this invention differs from traditional fixed-mode control systems, achieving collaborative optimization control under multimodal drive.

[0088] Example 4: Cooperative Control Strategy like Figure 3 As shown, this embodiment provides a detailed explanation of the collaborative control logic between deep peak shaving and primary frequency regulation. The core of this invention lies in achieving coordination and unification between deep peak shaving operation and primary frequency regulation response by employing different control strategies at different operating stages through a hierarchical collaborative optimization decision module 30, thus avoiding the problem of mutual interference between the two control objectives under traditional control methods.

[0089] I. Control Strategies for Deep Peak Shaving Operation When the operating mode recognition module 20 identifies that the unit is in a deep peak-shaving steady-state mode, the hierarchical collaborative optimization decision-making module 30 prioritizes economic operation and frequency regulation capacity reserve while ensuring the safe operation of the unit.

[0090] In this embodiment, the control strategy for the deep peak shaving stage includes: 1. Optimize the turbine control valve opening to keep the control valve within the operating range suitable for rapid adjustment and avoid the control valve being at its extreme opening position; 2. Appropriately increase the main steam pressure setpoint to ensure the boiler system maintains a certain steam reserve capacity; 3. Under the premise of meeting environmental protection and combustion stability requirements, balance control of fuel quantity and air volume should be implemented.

[0091] Through the above measures, a controllable steam energy storage reserve is formed under deep peak shaving steady state, providing a basis for rapid power release for subsequent frequency regulation operations.

[0092] In this embodiment, in order to quantitatively describe the steam energy storage capacity formed during the deep peak shaving stage, the hierarchical collaborative optimization decision module 30 establishes a dynamic estimation model for steam energy storage, which is used to evaluate the steam energy reserves that can be released under the current operating conditions.

[0093] Let E be the equivalent energy storage capacity of the main steam system of the unit, then it can be expressed as: E = C·(P-P0); in: E represents the energy stored in the currently available steam. C is the equivalent steam volume coefficient of the main steam system, which is related to the boiler steam volume, superheater volume and system pressure characteristics. P represents the current main steam pressure; P0 is the reference main steam pressure during deep peak shaving steady-state operation.

[0094] When P is greater than P0, it indicates that the system is in an energy storage state; when frequency regulation is triggered, the steam energy storage E can be used to support the rapid power output of the steam turbine.

[0095] Furthermore, to account for the impact of dynamic changes in steam pressure on available energy storage, a time correction can be applied to the steam energy storage capacity, resulting in: E(t) = C·(P(t) - P0); Where P(t) is a function of the main steam pressure as a function of time.

[0096] During a frequency regulation response, the adaptive control unit 32 dynamically adjusts the rate of change of the turbine valve opening and the boiler fuel compensation rhythm based on the steam energy storage estimation result E(t), so that the steam energy storage is released first during the rapid response phase of the turbine, and the combustion system follows up to restore energy balance.

[0097] This dynamic estimation model for steam storage makes the pressure increase during the deep peak shaving stage no longer just an empirical adjustment, but a quantifiable adjustment variable that can participate in control decisions, thereby enhancing the unit's primary frequency regulation capability under low load operating conditions.

[0098] The steam energy storage estimation result can be used as a feedforward input to the adaptive control unit to participate in the control quantity calculation, so as to improve the control accuracy in the frequency modulation response stage.

[0099] II. Control Strategy for the Primary Frequency Modulation Triggering Phase When the detected grid frequency deviation exceeds the preset frequency regulation trigger threshold, the operating mode identification module 20 switches the unit status to the primary frequency regulation response mode. At this time, the hierarchical collaborative optimization decision module 30 elevates the frequency regulation performance target to the highest priority.

[0100] In this embodiment, the following cooperative control strategy is adopted in the primary frequency modulation response stage: 1. Rapid regulation is prioritized by the turbine control valve, which adjusts the active power output quickly by changing the steam flow rate; 2. The boiler system utilizes the steam energy storage formed in the early stage to compensate and slow down the drop in main steam pressure; 3. Fuel quantity and air volume are adjusted in tandem over subsequent timescales to restore boiler energy balance.

[0101] Through a multi-stage coordinated control approach of "turbine rapid response - boiler energy storage compensation - combustion system follow-up", the unit can complete frequency support within a second-level time scale, while avoiding drastic fluctuations in main steam pressure.

[0102] III. Suppression of Reverse Coupling Regulation Mechanism In traditional control systems, when the turbine control valve opens rapidly, the main steam pressure drops, which may trigger the boiler main control to increase fuel in the opposite direction, resulting in control internal friction or even oscillation.

[0103] To avoid the above problems, in this embodiment, the hierarchical collaborative optimization decision module 30 temporarily adjusts the boiler pressure control strategy during the first frequency regulation response stage, so that the boiler main controller prioritizes maintaining pressure stability in a short period of time, rather than immediately making large compensation.

[0104] This coupling suppression mechanism reduces reverse regulation behavior between the boiler and the turbine, thereby improving the dynamic stability of the system.

[0105] IV. Recovery Mechanism After Frequency Modulation Ends When the grid frequency returns to the allowable range, the operating mode identification module 20 will gradually switch the unit status back to the deep peak shaving steady-state mode.

[0106] In this embodiment, the recovery process includes: 1. Smoothly adjust the turbine control valve opening to the deep peak shaving operation setting value; 2. Restore the main steam pressure to the deep peak-shaving steady-state target value; 3. Gradually restore the weight of economic priority control targets.

[0107] By setting a smooth transition mechanism, we can avoid oscillations in control parameters caused by frequent switching.

[0108] V. Comprehensive Technical Effects of Collaborative Control Through the above-described deep peak shaving and primary frequency modulation coordinated control strategy, this embodiment can achieve the following: 1. Maintain a certain reserve of frequency regulation capacity under deep peak shaving and low load operation conditions; 2. Achieve rapid response and controlled fluctuations during a single frequency modulation trigger; 3. Reduce drastic fluctuations in main steam pressure and temperature; 4. Avoid reverse coupling oscillations between the boiler and the steam turbine; 5. Quickly restore to economic operating conditions after the frequency adjustment ends.

[0109] Therefore, this embodiment achieves time scale coordination between minute-level peak shaving control and second-level frequency modulation control, effectively solving the inherent conflict between deep peak shaving and primary frequency modulation.

[0110] Example 5: Safety Priority Control like Figure 1 As shown, the safety protection module 50 and the collaborative execution module 40 are connected in parallel and interact with the hierarchical collaborative optimization decision module 30 to perform boundary monitoring of key parameters during unit operation and to activate the safety priority control strategy when necessary to ensure the reliability and stability of unit operation.

[0111] This embodiment mainly describes the structure and control logic of the safety protection module 50.

[0112] I. Safety Monitoring Targets In this embodiment, the safety protection module 50 focuses on monitoring the following key operating parameters: 1. Main steam pressure and its rate of change; 2. Main steam temperature and reheat steam temperature; 3. Negative pressure in the furnace; 4. Turbine control valve opening; 5. Relevant indicators of boiler combustion stability.

[0113] All of the above parameters have preset operating boundaries, including upper limit, lower limit and change rate limit values.

[0114] II. Safety Margin Calculation Method The safety protection module 50 calculates the margin value between each key parameter and its corresponding preset operating boundary in real time.

[0115] For example: The main steam pressure margin is the difference between the current main steam pressure and the minimum allowable pressure; The main steam temperature margin is the difference between the current temperature and the maximum allowable temperature; The furnace negative pressure margin is the distance between the current negative pressure and the set safe range.

[0116] When the margin value of any key parameter is less than the preset safety threshold, the safety protection module 50 enters the warning state; when the margin value is further reduced to the trigger threshold, the safety priority control strategy is activated.

[0117] III. Safety-First Control Strategy In this embodiment, the security priority control strategy includes the following steps: 1. Pause the current adjustment of the optimization target weights in the hierarchical collaborative optimization decision module 30; 2. Limit the rate of change of control commands to avoid large fluctuations in fuel quantity or valve opening; 3. Prioritize stable control of main steam pressure or main steam temperature; 4. Reduce the rate of change of unit load when necessary.

[0118] During periods of safety-priority control, the system prioritizes maintaining the stability of key parameters, while temporarily reducing the weight of economic or frequency regulation performance objectives.

[0119] IV. Security Recovery Mechanism When the critical operating parameters are detected to have returned to the safe range and continue to meet the set time conditions, the safety protection module 50 sends a recovery signal to the hierarchical collaborative optimization decision module 30.

[0120] The system then gradually restores the original optimized target weight configuration and smoothly transitions back to the normal collaborative control mode to avoid new fluctuations caused by sudden changes.

[0121] V. The Relationship Between Security Protection and Collaborative Control In this embodiment, the safety protection module 50 is not a protection device independent of the collaborative control, but rather an embedded part of the overall architecture of the control system.

[0122] Its functions include: 1. Provides operational boundary constraint information for the hierarchical collaborative optimization decision module 30; 2. Constrain the control strategy when key parameters approach their boundaries; 3. Improve the robustness of the system under complex operating conditions.

[0123] Through the above mechanism, the control system of the present invention can achieve deep peak shaving and primary frequency modulation coordinated optimization while having reliable safety assurance capabilities, avoiding operational risks caused by optimized control.

[0124] Example 6: Online Participation Method of Digital Twin like Figure 4 As shown, the digital twin support platform 60 is an important component of the control system of this invention. It is used to construct the unit mechanism model and data fusion model, and provides model parameters, operating boundary constraint information and dynamic prediction results to the hierarchical collaborative optimization decision module 30 through online interaction, thereby enhancing the control system's adaptability to complex operating conditions.

[0125] This embodiment describes the structural composition of the digital twin support platform 60 and its online participation methods.

[0126] I. Structural Components of the Digital Twin Support Platform In this embodiment, the digital twin support platform 60 includes: 1. Mechanism model unit 61; 2. Data-driven model unit 62; 3. Simulation verification unit 63.

[0127] (1) Mechanism model unit 61 Mechanism model unit 61 is used to establish a thermodynamic and dynamic behavior model of the boiler-turbine-generator system.

[0128] The mechanism model includes: Dynamic model of boiler combustion and steam generation; Model for the transfer of main steam pressure and main steam temperature; Model of the relationship between steam flow rate and power output of steam turbine; Unit energy balance model.

[0129] The above model can be used to describe the dynamic response characteristics of the unit under different loads and different control commands.

[0130] (2) Data-driven model unit 62 The data-driven model unit 62 is trained based on historical operating data to establish a dynamic prediction model or control compensation model for the unit.

[0131] The data-driven model can be used for: Predict the trends in main steam pressure and main steam temperature; Predict the impact of load changes on key parameters; Compensation is provided for nonlinear behaviors not covered by the mechanistic model.

[0132] Combining data-driven models with mechanistic models forms a mechanistic-data fusion model, which improves prediction accuracy.

[0133] (3) Simulation verification unit 63 Simulation verification unit 63 is used to simulate different operating scenarios in an offline environment, including: Deep peak shaving operation scenarios; A single frequency modulation disturbance scenario; Scenarios with rapidly changing loads; Complex disturbance scenarios.

[0134] The simulation verification unit 63 can be used to test and optimize the control strategy and conduct risk assessments before actual operation.

[0135] II. Online Participation Methods of the Digital Twin Platform In this embodiment, the digital twin support platform 60 is not only used for offline simulation, but also interacts with the hierarchical collaborative optimization decision module 30 in real time through an online interface.

[0136] Its online participation methods include: 1. Provide model parameter update information to the hierarchical collaborative optimization decision module 30 so that the controller parameters can be adjusted according to changes in the group state; 2. Provide short-term forecasts of key operating parameters to assist the controller in performing feedforward compensation; 3. Provides operational boundary constraint information to assist the safety protection module 50 in determining the safety margin.

[0137] Through the aforementioned online participation mechanism, the digital twin platform becomes a decision support unit for the control system, rather than an independent simulation tool. The prediction results generated by the digital twin support platform participate in the control quantity calculation process of the adaptive control unit, thereby affecting the generation of coordinated control commands between the boiler and the steam turbine.

[0138] In this embodiment, the prediction results generated by the digital twin support platform not only exist as reference information, but also participate in the control quantity calculation process of the adaptive control unit.

[0139] Specifically, when generating coordinated control commands for the boiler and turbine, the adaptive control unit incorporates the predicted values ​​of key parameters, dynamic response trends, or operating boundary constraint information output by the digital twin support platform as feedforward compensation or constraint correction terms into the control calculation model, thereby making real-time corrections to the control quantities.

[0140] For example, when the digital twin support platform predicts that the main steam pressure will show a downward trend in the near future, the adaptive control unit can adjust the fuel quantity or valve opening change rate in advance to suppress pressure fluctuations; when the predicted key operating parameters are close to the safety boundary, the safety stability weight in the optimization objective function can be dynamically increased.

[0141] In this way, the digital twin support platform and the control system are coupled in a computational relationship, rather than being an independently operating simulation tool.

[0142] III. The Synergistic Relationship Between Digital Twins and Control Closed Loops In this embodiment, the digital twin support platform 60 and the control system form the following closed-loop relationship: 1. Actual unit operating data is input into the digital twin platform; 2. The digital twin platform generates prediction results based on a mechanism-data fusion model; 3. The prediction results are fed back to the hierarchical collaborative optimization decision-making module 30; 4. The hierarchical collaborative optimization decision-making module 30 optimizes control commands based on the prediction results; 5. Control commands are applied to the actual generating units; 6. The new operational data is fed back to the digital twin platform.

[0143] The closed-loop structure described above enables the control system to have certain predictive and adaptive capabilities, thereby improving the stability and robustness of the unit under wide load ranges and complex disturbance conditions.

[0144] IV. Technical Effects By introducing the digital twin support platform 60, this embodiment achieves the following technical effects: 1. Improve the adaptability of the control system to complex nonlinear operating conditions; 2. Predict the changing trends of key parameters in advance to reduce the risk of exceeding limits; 3. Reduce the operational risks of control strategies; 4. Enhance the long-term stability of the system.

[0145] Therefore, the digital twin support platform 60 in this embodiment becomes an important technical support for the realization of multimodal adaptive collaborative optimization control in this invention.

[0146] Example 7: Frequency regulation process under 30% load To further illustrate the application of the control system of the present invention in actual operation, this embodiment combines a typical operating scenario to explain the entire process of the system participating in primary frequency regulation under a deep peak shaving condition of 30% rated load.

[0147] This embodiment is for illustrative purposes only and does not constitute a limitation on the scope of protection of this invention.

[0148] I. Initial Operating Status In this embodiment, the unit has a rated capacity of 600MW and the current operating load is 30% of the rated load, which is in a deep peak-shaving steady-state operation mode.

[0149] at this time: Main steam pressure is maintained at the deep peak shaving setpoint; The turbine control valves are kept at a medium opening range suitable for rapid adjustment; The boiler fuel and air volume are in a low-load, stable combustion state. The power grid frequency is near the rated frequency, and the frequency regulation response condition has not been triggered.

[0150] The mode recognition module 20 identifies the current state as "deep peak shaving steady-state mode" and sends this mode information to the hierarchical collaborative optimization decision module 30.

[0151] The hierarchical collaborative optimization decision-making module 30 sets economic operation and frequency regulation capacity maintenance as the main control objectives, while forming a certain steam energy storage reserve by appropriately increasing the main steam pressure setpoint.

[0152] II. Frequency regulation triggered by power grid frequency drop At a certain moment, the power grid frequency dropped from 50Hz to 49.90Hz, and the frequency deviation exceeded the preset frequency modulation trigger threshold.

[0153] The multi-source operation data sensing module 10 detects frequency deviation and frequency change rate signals and transmits the relevant data to the operation mode recognition module 20.

[0154] Based on the frequency deviation signal, the operation mode recognition module 20 determines that the unit has entered the "primary frequency regulation response mode" and sends the mode switching signal to the hierarchical collaborative optimization decision module 30.

[0155] III. Frequency Modulation Response Stage After receiving a frequency modulation response mode signal, the hierarchical collaborative optimization decision module 30 sets the frequency modulation performance target to the highest priority and generates a coordinated control command.

[0156] The specific process is as follows: 1. The adaptive control unit 32 first generates a command to quickly open the turbine control valve, thereby increasing the steam flow and rapidly improving the unit's active power output; 2. At the same time, the boiler system utilizes the steam energy storage formed during the deep peak shaving phase to buffer the drop in main steam pressure; 3. The fuel and air volume are gradually increased over the next few seconds to restore the boiler's energy balance; 4. The digital twin support platform 60 generates short-term prediction results based on real-time operating data, assisting the hierarchical collaborative optimization decision module 30 in fine-tuning control commands to avoid excessive fluctuations in main steam pressure; 5. The safety protection module 50 monitors the main steam pressure and temperature margin in real time to ensure that key parameters do not exceed preset boundaries.

[0157] Through the aforementioned coordinated control process, the unit completes its response to grid frequency deviation within seconds and gradually stabilizes at the new power output level.

[0158] IV. Frequency Modulation End and Recovery Process When the grid frequency returns to the normal range, the operating mode identification module 20 determines the unit status as deep peak shaving steady-state mode.

[0159] The hierarchical collaborative optimization decision module 30 gradually reduces the weight of the frequency modulation performance target and restores the priority of economic operation based on the modal changes.

[0160] The recovery process includes: 1. Smoothly adjust the turbine control valve opening to the deep peak-shaving steady-state setpoint; 2. Gradually restore the main steam pressure to the deep peak-shaving steady-state target; 3. Adjust the fuel quantity and air volume to a stable combustion state under low load.

[0161] The entire recovery process employs a gradual adjustment approach to avoid oscillations caused by sudden parameter changes.

[0162] V. Demonstration of Technical Effects As can be seen from the above operating scenarios, under the condition of deep peak shaving at 30% rated load, the control system of this invention can: 1. Maintain a certain level of steam energy storage reserves to improve frequency regulation response capability; 2. Achieve rapid and stable power regulation when the grid frequency deviates; 3. Reduce drastic fluctuations in main steam pressure and temperature; 4. Avoid reverse coupling regulation between the boiler and the steam turbine; 5. Quickly restore to economic operating conditions after the frequency adjustment ends.

[0163] Therefore, this invention effectively improves the primary frequency regulation performance of the unit under deep peak shaving conditions and achieves coordination and unity between deep peak shaving and primary frequency regulation.

[0164] In other embodiments of the present invention, the specific implementation of the above-mentioned operating mode recognition module, hierarchical collaborative optimization decision module and adaptive control unit is not limited to the structural form described in the foregoing embodiments.

[0165] For example, in addition to using cluster analysis combined with time-series feature recognition, the operation mode recognition module can also use rule-based judgment methods, statistical analysis methods, or other methods based on operation feature division. As long as it can classify and identify operation modes according to the unit's operating status, it falls within the protection scope of this invention.

[0166] Furthermore, in addition to the control method that combines the mechanism model and the data-driven model, the adaptive control unit can also be implemented using other control methods with predictive or adaptive capabilities, such as model predictive control based on parameter self-tuning mechanism or control strategy based on online correction mechanism. As long as the control objective can be dynamically adjusted according to the operating mode and coordinated control commands for boiler and turbine can be generated, they should all be regarded as equivalent alternatives of the present invention.

[0167] Furthermore, the digital twin support platform can be implemented using different model building methods, including but not limited to full-mechanism modeling, data-driven modeling, or mechanism-data fusion modeling. As long as it can provide prediction results, model parameters, or operational constraint information to the hierarchical collaborative optimization decision-making module and participate in online control decision-making, it falls within the protection scope of this invention.

[0168] The above-described alternative embodiments demonstrate that the technical solution of the present invention is not limited to a specific algorithm structure or model form, thereby enhancing the stability of the protection scope and its resistance to circumvention.

[0169] In a further embodiment of the present invention, the control objectives involved in the hierarchical collaborative optimization decision-making module include not only economic objectives, frequency regulation performance objectives, and safety and stability objectives, but can also be extended to environmental protection operation objectives.

[0170] Specifically, the environmental protection operation objectives include, but are not limited to: Nitrogen oxide (NOx) emission targets; Sulfur dioxide (SO2) emission targets; Control indicators for oxygen content in flue gas; Optimization indicators for the operating load of denitrification or desulfurization systems.

[0171] In this embodiment, the target weight dynamic adjustment unit can incorporate environmental protection operation targets into the control target weight system according to the operating mode and environmental protection constraints, so as to achieve synergistic optimization of pollutant emission control while meeting the performance requirements of deep peak shaving and primary frequency regulation.

[0172] For example, in the deep peak-shaving steady-state mode, priority can be given to ensuring that NOx emissions meet the standards; in the primary frequency regulation response mode, while ensuring a rapid frequency response, the combustion adjustment range is constrained to avoid instantaneous fluctuations in emissions.

[0173] By incorporating environmental protection operation objectives into a multi-objective collaborative optimization framework, the control system of this invention can adapt to more stringent environmental regulatory requirements and improve the overall performance of unit operation.

[0174] Compared with the prior art, the present invention has at least the following beneficial effects: 1. By operating the mode recognition and control target weight dynamic reconstruction mechanism, the control system can automatically adjust the control center of gravity according to the unit operation stage, avoiding the problem of mutual constraint between deep peak shaving and primary frequency regulation targets under the traditional fixed control mode, and fundamentally alleviating the inherent conflict between the two types of control tasks.

[0175] 2. By proactively constructing controllable steam energy storage reserves during the deep peak shaving phase and prioritizing the release of steam energy storage during the primary frequency regulation response phase, time-scale coordination between minute-level peak shaving control and second-level frequency regulation control is achieved, thereby improving the frequency response capability under low-load operating conditions.

[0176] 3. By employing a coordinated control sequence of "rapid turbine response - boiler energy storage compensation - combustion system follow-up", the reverse coupling regulation between the boiler and the turbine is suppressed, the fluctuation range of main steam pressure and temperature is reduced, and the dynamic stability of the system is improved.

[0177] 4. By introducing a digital twin support platform and involving it in the calculation of control quantities, the control system is equipped with predictive and model adaptive capabilities, thereby improving control accuracy and reducing the risk of model mismatch.

[0178] 5. By setting an embedded safety priority control mechanism, the operating boundary is monitored in real time during the optimization control process. When the key parameters approach the safety threshold, the priority of the control target is automatically switched, ensuring the safety of unit operation while guaranteeing the regulation performance.

[0179] Therefore, this invention achieves coordinated and optimized control between deep peak shaving and primary frequency regulation, improving the regulation capability, operational stability, and safety of thermal power units under the background of high proportion of new energy access.

[0180] In summary, this invention, by constructing a closed-loop control architecture of "sensing-modal recognition-hierarchical decision-making-cooperative execution," achieves coordinated and optimized control between deep peak shaving and primary frequency regulation of thermal power units under wide load operating conditions. This improves the unit's regulation capability and operational stability in the context of a high proportion of new energy integration, and has good prospects for engineering applications.

[0181] It should be noted that the above description is merely a preferred embodiment of the present invention, used to illustrate the technical principles and implementation methods of the present invention, and not to limit the scope of protection of the present invention. For those skilled in the art, various modifications, equivalent substitutions, or improvements can be made to the technical solutions of the present invention without departing from the spirit and substance of the present invention, and all such modifications, equivalent substitutions, or improvements should fall within the scope of protection defined by the claims of the present invention. The scope of protection of the present invention is determined by the appended claims.

Claims

1. A multimodal adaptive learning-based deep peak shaving and primary frequency regulation coordinated optimization control system for thermal power units, characterized in that, include: (1) Multi-source operation data sensing module, used to collect the operating parameters of the boiler system, turbine system and generator system of thermal power unit, as well as the power grid frequency signal and dispatch command signal; (2) Operation mode identification module, used to analyze the unit's operating status based on the operating parameters and identify the unit's current operating mode, wherein the operating mode includes at least the deep peak shaving steady-state mode, the deep peak shaving transition mode, the primary frequency regulation response mode, and the conventional load operating mode; (3) A hierarchical collaborative optimization decision-making module, including a target weight dynamic adjustment unit and an adaptive control unit. The target weight dynamic adjustment unit is used to adjust the priority of the control target according to the current operating mode, and the adaptive control unit is used to generate coordinated control commands for the boiler and the steam turbine. (4) A collaborative execution module, used to apply the coordinated control commands to the fuel regulation circuit, air volume regulation circuit, feedwater regulation circuit and turbine control valve regulation circuit respectively; (5) A digital twin support platform is used to construct a unit mechanism model and a data fusion model, and to provide model parameters, operating boundary constraint information or dynamic prediction results to the hierarchical collaborative optimization decision module to assist online control decision-making. The prediction results generated by the digital twin support platform participate in the control quantity calculation of the adaptive control unit. The control system adopts a closed-loop architecture of "perception-modal recognition-hierarchical decision-making-cooperative execution". By switching the control target driven by the operating mode, it achieves the coordinated optimization of deep peak shaving operation and primary frequency regulation response.

2. The control system according to claim 1, characterized in that, The operating mode identification module classifies the operating status based on unit load level, main steam pressure, main steam temperature, grid frequency deviation, frequency change rate, and load change rate. The current operating mode and mode confidence are output by combining cluster analysis with time-series feature recognition.

3. The control system according to claim 1, characterized in that, The target weight dynamic adjustment unit adopts different control target priority configurations under different operating modes, wherein: Under deep peak shaving steady-state mode, the priority objectives are economical operation and maintaining frequency regulation capacity; In the first frequency modulation response mode, the priority objectives are fast frequency response and response accuracy; Under deep peak shaving transition mode, the stability of operating parameters and equipment safety are the priority objectives.

4. The control system according to claim 1, characterized in that, The adaptive control unit adopts a control method that combines a mechanistic model and a data-driven model. Among them, the mechanistic model is used to handle the multivariable coupling relationship and operational constraints of the boiler-turbine system. Data-driven models are predictive or strategy optimization models trained on historical operating data. They are used to predict the dynamic response of the unit or compensate for the control output, thereby improving the control accuracy and stability over a wide load range.

5. The control system according to claim 1, characterized in that, Under deep peak-shaving operation, the hierarchical collaborative optimization decision module optimizes the turbine control valve opening and appropriately increases the main steam pressure setpoint to form a controllable steam energy storage reserve. To enhance the unit's rapid power response capability during a single frequency regulation operation.

6. The control system according to claim 1, characterized in that, During a frequency regulation response, the hierarchical collaborative optimization decision-making module adopts a control strategy that prioritizes rapid adjustment of the turbine control valve, supplemented by boiler energy storage compensation, and coordinated follow-up of fuel and air volume. This is to reduce main steam pressure fluctuations and decrease reverse coupling regulation between the boiler and the turbine.

7. The control system according to claim 1, characterized in that, Also includes: The safety protection module is used to monitor the margin between the main steam pressure, main steam temperature, furnace negative pressure, and other key operating parameters and preset operating boundaries in real time. When the key operating parameters approach the preset operating boundary, the current optimization control target is paused and a safety priority control strategy is initiated to ensure the safe operation of the unit.

8. A control method based on the control system according to any one of claims 1-7, characterized in that, include: S1: Collects unit operating parameters and grid frequency signals; S2: Identify the current operating mode; S3: Adjust the priority of control targets based on operating modes; S4: Generate coordinated control commands for boiler and steam turbine; S5: Execute the control command; S6: Assist in optimizing control strategies using model parameters or prediction results provided by the digital twin support platform; S7: Monitors key operating parameters in real time and initiates safety priority control when approaching the safety boundary.

Images